Note_Tech

All technological notes.

Project maintained by simonangel-fong Hosted on GitHub Pages — Theme by mattgraham

MCP

Model Context Protocol (MCP)

Model Context Protocol (MCP)
- a standardized way for AI models to interact with external systems such as tools, APIs, files, and databases.
- Think of MCP as “USB for AI” — a consistent interface that lets AI plug into different systems safely and predictably.

Key Components

Component	Description	Example
Host	The application the user interacts with; it runs and displays the AI	VS Code, chat app, web app
Client	The protocol layer inside the host that communicates with MCP servers	Python SDK client, MCP client library
MCP Server	A lightweight service that exposes tools, resources, and prompts	Backend service exposing APIs
Transport Layer	The communication method between client and server	HTTP, WebSocket
Tools	Functions the AI can call to perform tasks	Search flights, book ticket
Resources	Read-only data sources that provide context	Flight data, knowledge base, documents
Prompts	Predefined templates or instructions to guide AI behavior	“Summarize results”, “Explain options”
Schema	Defines how tools, resources, and prompts are structured and used	Tool input/output definitions (JSON format)

How MCP Works (Flow)

A user uses VS Code to ask an AI assistant to find flight tickets from Toronto to Vancouver.

User (Host) enters a prompt in VS Code
→ “Find me a flight from Toronto to Vancouver next weekend”
Client (AI / Agent) understands the request
→ Decides it needs flight data (external information)
Client → MCP Server (via Transport Layer)
→ Sends a request through HTTP/WebSocket to discover available tools
MCP Server responds with available capabilities
→ Tools: search flights, book flights
→ Resources: airline data, pricing info
→ Schema: how to call these tools
Client (AI) selects the appropriate tool
→ Chooses: “search flights”
Client → MCP Server (Tool Call via Transport Layer)
→ Sends structured request to search flights (based on schema)
MCP Server (Tools + Resources) executes the request
→ Fetches flight data from external systems
MCP Server → Client returns results
→ Flight options with prices and times
Client (AI) processes the result
→ Summarizes options into human-readable response
Client → User (Host / VS Code) displays the answer
→ “Here are the available flights…”

Lab: Kiro CLI with AWS MCP Server CloudFomation

Installation

ref: https://kiro.dev/docs/cli/
In Linux(no windows)

sudo apt update
sudo apt install unzip
curl -fsSL https://cli.kiro.dev/install | bash
# Kiro CLI installer:

# Downloading package...
# ✓ Downloaded and extracted
# ✓ Package installed successfully

# 🎉 Installation complete! Happy coding!

# 1. Important! Before you can continue, you must update your PATH to include:
#    /home/ubuntuadmin/.local/bin

# Add it to your PATH by adding this line to your shell configuration file:
#   export PATH="$HOME/.local/bin:$PATH"

# 2. Use the command "kiro-cli" to get started!
export PATH="$HOME/.local/bin:$PATH"

kiro-cli --version
# kiro-cli 1.29.8

mkdir myproject
cd myproject
kiro-cli

Install AWS CLI

Require AWS Cli

curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
sudo ./aws/install
aws --version
# aws-cli/2.34.29 Python/3.14.3 Linux/5.15.153.1-microsoft-standard-WSL2 exe/x86_64.ubuntu.24

aws configure

Steps

User(prompt) -> Kiro Cli(MCP Host and Client) -> MCP Server(AWS MCP Server CloudFomation) -> tools -> Resources(AWS CloudFormation)

Install MCP

ref: https://awslabs.github.io/mcp/servers/cfn-mcp-server

Install AWS IaC MCP Server - https://awslabs.github.io/mcp/servers/aws-iac-mcp-server

Create a bucket

Create an S3 bucket with versioning with the name - "test_mcp_5f85j3f"

Create a hello world web app

Create terraform files at the path web_app/ for a hello world html web app hosted on the bucket named web_hello_world_8fj40f.

deploy this app